Autism Spectrum Disorder (ASD) is a neurodevelopmental condition that affects social communication, behavior, and sensory processing. Traditional diagnostic methods rely on clinical observations, which are time-consuming, subjective, and prone to inconsistencies. This study explores automated ASD detection using two distinct modalities: Eye-tracking images and electroencephalogram (EEG) signals, employing machine learning and deep learning techniques. Initially, we investigated eye-tracking data, analyzing gaze patterns with pre-trained deep learning models such as VGG16, ResNet, InceptionV3, and EfficientNet. Despite rigorous experimentation, the highest achieved accuracy remained 82% (VGG16). To improve performance, we developed a custom CNN model for eye-tracking images with three convolutional layers, max pooling, and fully connected dense layers, trained using the Adam optimizer and Binary Crossentropy loss. However, this custom model achieved only 72.73% accuracy, further confirming the limitations of eye-tracking data for ASD detection. Recognizing these challenges, we shifted our focus to EEG-based classification, leveraging advanced signal processing techniques including Independent Component Analysis (ICA), Power Spectral Density (PSD) analysis, and Fast Fourier Transform (FFT) for feature extraction. Our custom deep learning CNN model for EEG signals significantly outperformed image-based classification, achieving 87.89% accuracy, making EEG a more reliable biomarker for ASD detection. This study demonstrates the superiority of EEG-based classification over eye-tracking images and contributes to the field of biomarker-based ASD diagnosis. Future work will explore multi-modal fusion of eye-tracking and EEG data to enhance classification performance and support early, automated ASD screening.